LocalLLaMA
Welcome to LocalLLaMA! Here we discuss running and developing machine learning models at home. Lets explore cutting edge open source neural network technology together.
Get support from the community! Ask questions, share prompts, discuss benchmarks, get hyped at the latest and greatest model releases! Enjoy talking about our awesome hobby.
As ambassadors of the self-hosting machine learning community, we strive to support each other and share our enthusiasm in a positive constructive way.
Rules:
Rule 1 - No harassment or personal character attacks of community members. I.E no namecalling, no generalizing entire groups of people that make up our community, no baseless personal insults.
Rule 2 - No comparing artificial intelligence/machine learning models to cryptocurrency. I.E no comparing the usefulness of models to that of NFTs, no comparing the resource usage required to train a model is anything close to maintaining a blockchain/ mining for crypto, no implying its just a fad/bubble that will leave people with nothing of value when it burst.
Rule 3 - No comparing artificial intelligence/machine learning to simple text prediction algorithms. I.E statements such as "llms are basically just simple text predictions like what your phone keyboard autocorrect uses, and they're still using the same algorithms since <over 10 years ago>.
Rule 4 - No implying that models are devoid of purpose or potential for enriching peoples lives.
view the rest of the comments
I used it quite a lot at the start of the year, for software architecture and development. But the number of areas where it was useful were so small, and running it locally is quite slow. (which I do for privacy reasons)
I noticed that much of what was generated needed to be double checked, and were sometimes just wrong, so I've basically stopped using it.
Now I'm hopeful for better code generation models, and will spend the fall building a framework around a local model. See if the helps in guiding the models generation.
True. I didn't even bother giving it tasks like that. I don't think AI is going to replace software designers or programmers anytime soon. Well... maybe except for simple stuff, copy-paste programming, simple scripting, webdesign and some self contained and not overly complex stuff. It's a fascinating tool, can help you do things quickly, answer questions, do prototypes. But if you throw real work at it, you soon realise it has severe limitations and isn't even close to the human intellect. Okay. Maybe you consider it heaven-sent if you studied history instead of computer science and it provides you with python-scripts to sort your data.
Journalists and people who write text have similar problems. The chatbots generate convincing text and can take over some of the work writing text. But for example if you need to write a text that is correct and factual, you'd be better off without AI. At least that's what i read in some articles about ChatGPT. Everyone needs to put in considerable effort to fact check its output, and double check everything to the point that it doesn't make sense to run the AI in the first place.
I learned the hard way to never generate anything I couldn't create myself, of at least verify its validity.
I'm pumped for Llama2 which was released yesterday. Early tests slow some big improvements. Can't wait for Wizard/Vicuna/Uncensored versions of it.
Yeah, me too. After summer vacation it's hopefully available and I'll dig into it.
It’s marginally better than original but WAYY more censored. It is pretty intrusive. It refused to write a bash script to kill a process by regexp 🤦
The first uncensored variants are already on Huggingface though, look for The Bloke. :)
I just watched the youtube video that got linked here earlier. I forgot if it was better or worse at programming than its predecessor. but it's not that much a difference. i'm just now fiddling around with the chat variant. But i'm excited for the tuned versions, too. (thrilled)